The step 3 -> 4 is jarring on Chrome 48 OSX.
I don't believe this format works well with a scrolling page to begin with. It makes a lot more sense on full-screen apps, or on fixed navigation menus. If you really need to show a dialog further down the page, you're better off having the trigger for it be down there - not further up the page.
Most importantly, don't overuse it. Even the most complex situation should not need more than 2-3 of these. Beyond pointing out extremely critical elements to interact with, the page or app shouldn't be that difficult to navigate.